Augmentative Topology Agents For Open-Ended Learning

TL;DR

This paper introduces ATEP, an approach that evolves neural network topologies of agents in tandem with challenging environments, leading to more adaptable and capable agents in open-ended learning scenarios.

Contribution

The paper presents ATEP, a novel method that allows agents to evolve their neural network structures dynamically, enhancing generalization in open-ended learning environments.

Findings

ATEP produces agents that solve more environments than fixed-topology baselines.

Evolving network complexity improves agent adaptability and performance.

Species-based transfer mechanisms further enhance generalization.

Abstract

In this work, we tackle the problem of open-ended learning by introducing a method that simultaneously evolves agents and increasingly challenging environments. Unlike previous open-ended approaches that optimize agents using a fixed neural network topology, we hypothesize that generalization can be improved by allowing agents' controllers to become more complex as they encounter more difficult environments. Our method, Augmentative Topology EPOET (ATEP), extends the Enhanced Paired Open-Ended Trailblazer (EPOET) algorithm by allowing agents to evolve their own neural network structures over time, adding complexity and capacity as necessary. Empirical results demonstrate that ATEP results in general agents capable of solving more environments than a fixed-topology baseline. We also investigate mechanisms for transferring agents between environments and find that a species-based approach further improves the performance and generalization of agents.